I have been downloading the daily average global surface air temperature data initial condition output for the Sigma 0.995 level from the National Centers for Environmental Prediction and National Center for Atmospheric Research Reanalysis 1 (NCAR R1) cooperative effort. This output is still being updated daily about 2 days behind the current day and extends back to 1948. So far I have downloaded and processed the daily temperature output back to 1979. The Sigma 0.995 level corresponds to the pressure altitude at 99.5 % of the surface air pressure, which is roughly about 50 meters above ground level. The actual height above ground level varies somewhat depending on atmospheric conditions. The NCAR R1 model uses a 2.5 degree latitude by 2.5 degree longitude grid. In this post I am comparing the resulting global mean surface air temperature anomalies (GMSATA) for several time periods and different reference baseline periods. See the Methods section at the end for calculation details and links to sources.
The first two graphs, Figures 1 and 2, display NCAR R1 daily average GMSATA time series output for two different reference baselines. I was expecting the older 1979-2000 baseline to show higher temperature anomalies than the much more recent 2011-2015 baseline as can be seen in the graphs, but I was surprised that the baseline shift also changed the seasonal pattern across each year. Apparently the seasonal patterns have shifted from one baseline period to another. Figure 3 is provided to compare the daily average 2-meter above ground level surface air temperature anomalies from the more modern Climate Forecast System Reanalysis (CFSR) to the NCAR R1 output. The CFSR has a higher resolution of 0.5 degree latitude by 0.5 degree longitude. See the Daily Updates page, which can be accessed from the menu bar at the top of this page, for more information about the daily CFSR output.
I also calculated GMSATA for two other reference baseline periods, 1994-2013 and 1981-2010. The time series results for all four reference baseline periods for 2018-2019 are plotted together in Figure 4. They all converge around January-February and diverge the most around September-October. I’m not sure why.
I included the 1994-2013 reference period used by Nick Stokes for reporting daily averages here and the 1979-2000 period used by the University of Maine Climate Reanalyzer here. The 1981-2010 period is the most recent three decade (30-year) period commonly used for climatological data reporting. I previously used the 2011-2015 period for comparing monthly NCAR R1 versus CFSR temperature anomalies here.
The next three graphs, Figures 5 through 7, are like the first three graphs, but covering a longer time period, from 2014 to 2019 so far. Again the general patterns are similar, but the details differ.
The last two graphs in Figures 8 and 9 cover a longer time period, for the current century so far, beginning in 2001. I do not yet have CFSR daily averages for all of this period, so only the NCAR R1 results for two different reference periods are presented.
Overall, this exercise goes to show that changing reference baseline periods for daily GMSATA does cause quite a bit of variation in the results – more than I expected. However, the general trends as indicated by the running 365-day averages did not appear to be affected by changing baselines.
For the NCAR R1 daily averages I downloaded the gridded Sigma 0.995 level temperature output which is provided in compacted annual files (thanks to Nick Stokes for providing the link below on his blog). I used the National Aeronautic and Space Administration (NASA) Panopoly program to extract the temperature grids from the compacted data files and then loaded the daily temperature grid data into spreadsheets by year. For each day I calculated daily averages by latitude zone, weighted by area, to compile zonal and global averages.
I calculated the reference period baseline averages for each day and used centered running 5-day averages to smooth the baseline results. Once the reference baseline temperatures were calculated, the temperature anomalies were calculated for each day by subtracting the reference baseline value for the day from the actual daily temperature average for that day. For the temperature anomalies I also calculated running 365-day and 91-day averages to show annual and seasonal scale tendencies.
NCAR R1 annual compacted files with daily global temperature grids:
NASA Panoply program: